Open YaroShkvorets opened 1 year ago
This seems to happen at shutdown of the whole stack, so I'll take a closer look.
The code around shutdown has been heavily refactored in v2 bump, I'll run again with race detection to see if it stills report the problem.
Ok it actually finds other also now. I'll try to take time next week to dig deeper.
Looks like there is a potential data race scenario that could result in the sink writing incorrect cursor to the database. Hard to recreate but race detecter sees it: